Improving the Diagnosis of Phenylketonuria by Using a Machine Learning–Based Screening Model of Neonatal MRM Data

Abstract

Phenylketonuria (PKU) is a common genetic metabolic disorder that affects the infant's nerve development and manifests as abnormal behavior and developmental delay as the child grows. Currently, a triple-quadrupole mass spectrometer (TQ-MS) is a common high-accuracy clinical PKU screening method. However, there is high false-positive rate associated with this modality, and its reduction can provide a diagnostic and economic benefit to both pediatric patients and health providers. Machine learning methods have the advantage of utilizing high-dimensional and complex features, which can be obtained from the patient's metabolic patterns and interrogated for clinically relevant knowledge. In this study, using TQ-MS screening data of more than 600,000 patients collected at the Newborn Screening Center of Shanghai Children's Hospital, we derived a dataset containing 256 PKU-suspected cases. We then developed a machine learning logistic regression analysis model with the aim to minimize false-positive rates in the results of the initial PKU test. The model attained a 95-100% sensitivity, the specificity was improved 53.14%, and positive predictive value increased from 19.14 to 32.16%. Our study shows that machine learning models may be used as a pediatric diagnosis aid tool to reduce the number of suspected cases and to help eliminate patient recall. Our study can serve as a future reference for the selection and evaluation of computational screening methods.

Opportunities and challenges in machine learning‐based newborn screening—A systematic literature review

“…The main reasons for dismissing papers were that they did not apply ML classification methods, 28 investigated other diseases from NBS programs such as hearing disabilities 29 or did not use data obtained from MS/MS. 30 Publications from different screening centers in Europe, [12][13][14][15][16]19 Asia, 7,17,18,20,21,23,24,26 and North America 22,25,27 are reviewed in this work.…”

“…Model agnostic pattern recognition can be applied for noninterpretable methods by discovering nonexplainable incidents such as a higher percentage of false positive newborns with Hispanic ethnicity 22,35 . This can be especially beneficial for varying prevalence between racial/ethnic groups and populations 25,26,35 . Furthermore, these methods can help to identify other risk factors such as gender , family disease history , and chronic diseases to identify infants with potential disease risk 24 …”

Inferring Time-Lagged Causality Using the Derivative of Single-Cell Expression

“…FN represents the number of edges in but not correctly directed in . These three metrics have been used in many studies [ 25 , 26 , 27 , 28 , 29 ]. In addition, we used the ability of determining directions (ADD) to evaluate how well a method was able to define directions given true causal edges.…”

Current Scenario and Future Direction of Newborn Screening and Management Program for Phenylketonuria in Bangladesh

“…However, most of the developing countries have focused on quality of health services and abundant evidence proposes that quality of health care is almost absent in case of inherent newborn deficiency disorders [2][3][4]. Although new-born screening may turn to the earliest detection of inherent disorders, some screening test results have not been provided to advantage the baby being screened; false diagnostic results may create adverse consequences to the baby and its parents [5].…”